GeForce RTX 3080 Performance on [email protected] CUDA-Enabled GPU Workloads
[email protected] GPU Workloads Become CUDA-Enabled!
The [email protected] project started the week by announcing that NVIDIA GPU users should soon be seeing a big boost in performance as their projects are now fully CUDA-enabled! NVIDIA engineers and the [email protected] team used the open source OpenMM toolkit to make this possible and are seeing 15-30% speedups on most [email protected] GPU projects. This drastically increasing both science throughput and points per day (PPD) these GPUs will generate.
This is a significant increase in performance, so after reading the blog post announcing this new feature we wanted to check it out. We also wanted to see how [email protected] performs on the new NVIDIA GeForce RTX 3080 graphics card as we haven’t run it on any Ampere GPU-based cards yet!
[email protected] is currently helping ambitious crowdsourced projects like the COVID Moonshot that is an open science drug discovery project looking to accelerate a proper working COVID antiviral. The project evaluates thousands of molecules per week with the hopes of producing a new low-cost patent-free therapy for COVID-19.
We tossed an ASUS TUF GAMING NVIDIA GeForce RTX 3080 video card ($699 shipped) into one of our Intel Core i7 test systems (Intel Core i7-8086K 6-core, 12-thread) and fired up the [email protected] client. The first workload that we got was running the 0.0.13 release of core22, which is the first CUDA-enabled version of the core.
Our logs showed that the CUDA runtime compiler libraries needed to accelerate the [email protected] were successfully downloaded by the client. The client logs also showed that the 0.0.13 core successfully launched the faster CUDA version. This is all done automatically, which is excellent!
We let it run for a bit on the ASUS TUF RTX3080 graphics card was was getting just over six million points per day (PPD) on the NVIDIA GeForce RTX 3080 graphics card. This was achieved by manually overclocking the card with the ASUS GPU Tweak II overclocking software utility. The Core clock was increased to 2000 MHz and the memory clock up to 19242 MHz.
The workload you are running has a big impact on your PPD, but this is impressive!
If you’d like to try out folding at home you can download the installer for Windows, macOS or Linux and get running in a couple minutes. Feel free to support Legit Reviews team #38296 if you try it out and choose COVID-19 if you’d like to help research that specific area. You can also choose to focus on research that is fighting Alzheimer’s, Cancer, Huntington’s, Parkinson’s or other research deemed high priority!